Can You Give Me Some Stats to Prove that Russell Westbrook Steals Rebounds?

          Properly valuing and evaluating rebounds has become something of a hot topic in recent years with Russell Westbrook averaging a triple double for three consecutive seasons despite being one of the smallest players in the history of the league to average 10 rebounds per game. Much of the discussion about his incredible feat has centered on whether or not it is the case that Westbrook’s rebounding numbers are inflated due to Westbrook taking a disproportionate amount of defensive rebounds which could be collected by other members of his team.

          Naturally, a lot of people are looking for stats to support the conclusion which they’ve already reached (thus the tongue-in-cheek title). I think we can analyze this question, within the context of a useful estimation of rebounding value for the entire league.

          Many people prefer to separate defensive rebounding from everything else that happens on defense. The same is of course true of offense as well; some would rather evaluate offense without taking into account the value of offensive rebounds. For those who prefer to consider rebounds as a separate category or phase of the game, this analysis will be right up your alley. In what follows, we are going to analyze the value added by rebounding independent of the rest of the game.

The Difference Between Offensive Rebounds and Defensive Rebounds

          When we consider this question (or the one about Russell Westbrook), it’s important to acknowledge a foundational difference between offensive and defensive rebounding. There is a deceptive balance between offensive rebounding and defensive rebounding: broadcasts and box scores compare team rebounds to determine who cleaned the glass better.

          The balance is illusory, however. The defense grabs approximately 76% of available rebounds. Quite often, the offense does not heavily contest a rebound, since doing so would incur the risk of giving up a fast break the other way if the offense sends everyone to the glass but fails to secure the rebound. In fact, the offense wins a much larger share of contested rebounds than uncontested rebounds (41-43% of all contested rebounds compared to 13-14% of uncontested rebounds in most seasons). This happens because it is rarely the case that a rebound can only be recovered by the offense without any defensive player being able to fight for the ball at all. Many of these rebounds are obviously the result of misses in transition on 2-on-1 or 3-on-1 breaks where the only defender goes after the shooter, thus leading to an uncontested offensive rebound.

Examples from 2018-19; note that the offense converted 42.5% of contested rebounds and 13.5% of uncontested rebounds in this season.

Valuing Defensive Rebounds

           All of this means that there is a difference in value when comparing contested rebounds with uncontested rebounds. For the defensive player, contested rebounds are a better indicator of rebounding performance than uncontested rebounds. Uncontested defensive rebounds are rebounds that the defensive team expects to convert a high percentage of – they represent plays where the ball can only be retrieved by one team (based on the location of ball and players). Unless the offense is unusually aggressive, the majority of these plays will result in defensive rebounds.

          More specifically, if we compare the inverse of the conversion rates for contested rebounds and uncontested rebounds, we find that uncontested rebounds are about 31% less difficult to convert for the defense than contested rebounds. For our purposes, that makes uncontested rebounds around 31% less valuable for evaluating a player’s rebounding performance. Thus, I weight rebounds appropriately such that contested rebounds are counted 1-for-1, while uncontested rebounds are multiplied by (one minus the percent change). Adding the two results (contested rebounds plus .69 times uncontested rebounds) gives us the player’s Rebounding Value Added for defensive rebounds.

          Since the defense expects to clean the glass and end the possession, we need to also account for the fact that they are not always successful. On the player level, that means we need to estimate how many rebounds a player has lost and not just how many rebounds he has gained.

          We can do this in a simple manner, by subtracting the rebounds grabbed by the player and those taken by one of his teammates from the player’s rebound chances. What is better than simply counting how many chances a player did not convert, however, is to use the data we have to make an estimate of how culpable he is for losing those rebounds.

Calculating Defensive Rebounds Lost

          Suppose that we begin by multiplying the player’s total rebounds lost while on defense (so offensive rebounds against him) by the percentage of offensive rebounds that are contested. League-wide, this figure is just a bit above 60% for the time period under consideration. We now have a preliminary estimate of how many of the rebounds that the player failed to recover were contested (and thus how many were uncontested).

AP Photo/Carlos Osorio

Next, we modify the preliminary estimate slightly by way of a measure I call Rebounding Toughness. A player’s Rebounding Toughness for defense is his contested defensive rebounds divided by his adjusted chances (chances minus rebounds that were grabbed by one of his teammates). The ratio tells us how many of the player’s rebound chances resulted in rebounds grabbed by that player which he had to fight for. There is, unsurprisingly, a significant difference in this figure for perimeter defenders as compared with interior defenders.

Comparing the player’s Rebounding Toughness with the average figure for his position gives us a multiplier slightly above or below 1.0 which we can use to adjust our initial estimate of the player’s contested rebounds lost. A player who is better at fighting for rebounds is a player whom we surmise to have given up more contested rebounds than expected and fewer uncontested rebounds than expected. If a guy who is good at getting his nose dirty loses a rebound, we assume that there must have been several opponents keeping him off the glass. On the contrary, a player who does not make an impact on the defensive glass probably gives up more uncontested rebounds than expected: he is out of position or gets pushed out of position.

In the table below, you can find the Defensive Rebounds Lost for any player in any season since 2015-16. The table’s color spectrum corresponds with DReb Lost per 48 Minutes (which you can also use as a filter).

Valuing Defensive Rebounds Lost

          To determine how to weight rebounds lost relative to one another (in other words, to figure out how much worse an uncontested rebound lost is than a contested rebound lost), we need to compare the value lost with the expected value. The value lost reflects the observed outcome (zero rebounds) minus the expected outcome (.86 rebound for uncontested rebounds, .59 rebound for contested rebounds). The difference expresses how much value a team loses for each type of rebound. Since both terms express value lost, both are negative numbers. Because uncontested rebounds lost forfeit more value than contested rebounds lost, the term for damage from uncontested rebounds lost is less than the term for damage from contested rebounds lost.

The question we need to address, though, is “how much worse?” Contested rebounds lost are 31% less damaging than uncontested rebounds lost, so my way of responding to this question is to weight uncontested rebounds lost at 100% and contested rebounds lost at 69%. The result is that a player whom we estimate to have lost a higher ratio of uncontested rebounds will grade out worse than a player whom we estimate to have lost a higher ratio of contested rebounds.

          Subtracting rebound value lost from rebound value added gives us a player’s Net Defensive Rebounding Value. You can evaluate these numbers per 48 minutes, compare them with position average, and all sorts of fun things. In the chart below, you can track Rebounding Value Added against Rebounds Lost. The players are color-coded according to their percent change relative to average for their position, meaning that a player far above average for his position is dark blue, and a player far below average for his position is dark orange. Note that although Rebounds Added and Lost are generally proportional to one another, player who are elite rebounders relative to their position may deviate substantially from the norm. (this chart is also available from the Stats menu)

Evaluating Offensive Rebounds

          The process for offensive rebounds gained and lost is almost identical, except in the rebounds lost phase. Whereas we used the player’s Rebounding Toughness to indicate how many of the rebounds that he lost were contested when we are evaluating defensive rebounds, we use the inverse of Rebounding Toughness to estimate how many of the rebounds that the player lost were contested on the offensive glass. The reason for this reversal is that losing contested rebounds is more damaging for the offense than losing uncontested rebounds; not contesting the rebound allows the offense to get back on defense, whereas crashing the glass puts the offense at risk of giving up easy transition opportunities the other way.

What losing a rebound looks like
Photo by Chris Humphreys-USA TODAY Sports

Losing a rebound when the offense is crashing the boards is thus a greater negative for the offense, since they lose possession of the ball and find themselves in suboptimal position following the possession change. In order to reflect this, we need to assign more contested rebounds lost to weak rebounders, and fewer contested rebounds lost to good rebounders. Think of it this way: if the offense is crashing the boards hard, but loses the rebound, it’s more likely that a weak rebounder got boxed out or had the rebound taken from him than that a strong rebounder got pushed out of position.

Weighting Offensive Rebounds Lost

          Because of this key difference, the weighting of rebounds on the offensive end has to be reversed as well. Since, as discussed above, the offense’s conversion rate is actually higher for contested rebounds than for uncontested rebounds, we cannot use the conversion rates to help us appropriately value uncontested rebounds compared with contested rebounds. In order to determine how much it costs a team to contest an offensive rebound that they don’t get, it helps to think about the expected value for the other team compared with the observed value.

When the offense contests a rebound, we expect the defense to be about 26.5% less effective at winning the rebound. Whenever the defense does win these rebounds, however, we actually observe a raw gain of 41 percentage points (1.00 rebound gained minus 0.59, the defense’s normal conversion rate on contested rebounds). When we compare the observed effect with the expected effect, we find that the defense gains 55% in value ((41.0-26.5)/26.5*100). From this we can conclude that losing a contested rebounds costs the offense 55% in expected value as compared with observed value.  

Having colorful hair helps you win contested rebounds

          When the offense loses an uncontested rebound, by contrast, we see a smaller difference between expected results and observed results. The defense gains one rebound, just as before. In this case, however, the defense already expected to win the rebound 85.5% of the time. So collecting the rebound represents only a 17% change in the defense’s fortunes, relative to their initial expected value. In simpler language, going from 85.5 to 100 is only a 17% increase (percent change = ((100 – 85.5)/85.5)*100). The expected value was already high, so most of the value of grabbing an uncontested rebound is expected value from the defense’s point of view.

          Now we know how much effect it has on the defense for the offense to surrender both contested and uncontested rebounds to them. Evaluating these two factors, we find that the value gained by the defense, and thus lost by the offense, is 69% less for uncontested rebounds than for contested rebounds. As a result, we multiply a player’s contested rebounds lost while on offense by 31%, then add them to the total number of the player’s uncontested rebounds lost while on offense.    

In the chart below, you can evaluate offensive rebounding performance for any player-season since 2015-16. As before, the players are placed on a color scale representing percent change relative to position. This helps to orient us when comparing smaller players against larger players. This chart is also available in the Stats section of this site.

But What About Russell Westbrook?

          In 2015-16, the year before all the triple double mania began and the final year that Westbrook played with Kevin Durant, Westbrook was an outstanding rebounder. His Net Offensive Rebounding Value was the 56th best in the league, but the most among offensive number 1’s. Westbrook’s Offensive Rebounding Net per 48 Minutes was also best among number 1’s, edging Kawhi Leonard by a microscopic seven hundredths of a rebound per 48. His Position-Adjusted Offensive Rebounding Value per 48 was 14th-highest in the league. So Russell Westbrook has always been a phenomenal offensive rebounder.

Brad Rempel-USA TODAY Sports

          In 15-16 Westbrook had the 27th-best Defensive Rebounding Value in the league, though he was only third in his position group (defensive number 2’s) behind Kawhi Leonard and Paul George. He was also third in his position group when evaluated per 48 minutes behind Leonard and Will Barton. Within the context of the whole league, though, Westbrook was only number 117 in Defensive Rebounding Value per 48. Westbrook’s position-adjusted value was third in the league behind the same two players – Kawhi Leonard and Will Barton. On the defensive glass, then, it looks as though Westbrook was among the best players at his position in 15-16, but not exactly among the best in the league.

          Westbrook’s Total Net Rebounding Value ranked number 41 in the league, though everyone ranked above him was taller than him. He was 140th in the league per 48 minutes, though only two players of remotely similar size finished above him (Kawhi Leonard and Tony Allen). His Position Adjusted Net per 48 Minutes was fifth in the league behind Andre Drummond, Zach Randolph, Kawhi Leonard, and Hassan Whiteside.

          Taking all the information together, it is fair to say that before Kevin Durant left the Thunder, Russell Westbrook was a fabulous rebounder overall. He was the best number 1 in the league on the offensive boards. On the defensive glass, he was the best small player in the league, though his totals were surpassed by a number of big men. Westbrook’s own build even served to dampen his incredible effectiveness for his size, as his ability to defend bigger guards as well as smaller ones led him to be categorized as a defensive 2 by my position algorithm. Compared with defensive 1’s, Westbrook would have far outstripped his peers (his figure was half a rebound higher than Marcus Smart and Ricky Rubio, the two highest defensive 1’s for the season).  

          He was not one of the most valuable rebounders in the league. As mentioned above, his total rebounding value was 41st in the league. Enes Kanter and Kevin Durant were the most productive rebounders on the 2015-16 Thunder, with Westbrook coming in third. The issue we want examine is: did this change in the last two seasons? Westbrook was already the best small rebounder in the league in 15-16. He bested other players of his size by a wide margin, even when Kevin Durant was still on his team. In the last three seasons, however, Russell Westbrook has been among the few players in the league to average 10 rebounds per game, besting even the majority of big men. What we want to know is whether he has actually become one of the top 10 rebounders in the league, or whether the totals overstate his value.

Is Westbrook Really One of the Best Rebounders in the League?

          In 2016-17, Russell Westbrook became the second player in NBA history to average a triple double for an entire season. He was once again a very good offensive rebounder. Westbrook ranked 45th in the league in Net Offensive Rebounding Value, and was third-highest among offensive number 1’s (behind Giannis Antetokounmpo and Jimmy Butler). He ranked 57th in the league in Position-Adjusted Offensive Rebounding Value per 48, down from 14th in 2015-16, so it is fair to say that his offensive rebounding performance declined slightly after taking on the extra offensive responsibility following the departure of Kevin Durant.

(AP Photo/Jack Dempsey)

          The difference clearly came on the defensive end, where Westbrook vaulted to fifth in the league in Defensive Rebounding Value, and eighth-best in the league per 48 minutes. He easily led the league in Position-Adjusted Defensive Rebounding Value per 48 Minutes.

          Westbrook’s Total Net Rounding Value was eighth-highest in the league, though he was only 17th-best per 48 minutes. He was 14th in the league per 48 minutes when accounting for position, though accounting for position actually allowed Enes Kanter to move in front of Westbrook as leader on the Thunder; this statement comes with the caveat that Russell bested Kanter in both Total Net Rebounding Value and Net per 48 Minutes.

          On the team level, Russell was third on his team in offensive rebounding behind Enes Kanter and Steven Adams, but easily led the Thunder in all defensive rebounding measures. Clearly, Westbrook increased his performance on the defensive glass in 2016-17, while maintaining a similar level of production on the offensive glass. As a result, he placed among the top 20 rebounders in the league in all total rebounding metrics. His rank of 11th in rebounds per game was apparently a relatively fair reflection of his actual value as a rebounder.

What About 2017-18?

          The 2017-18 season saw Westbrook average a triple double for the second consecutive season, an unprecedented achievement of such historical significance that it brought to a fever pitch the debate over whether or not Westbrook pads his stats. The debate dominated the NBA landscape in the leadup to the Thunder’s final game of the season, in which Westbrook needed 16 rebounds to put him above 10 rebounds per game for the season.

Westbrook’s value on the defensive glass was slightly lower in 2018 than in 2017, though still similar: he ranked fifth in the league in Net Defensive Rebounding Value (fifth in 2017), and only 23rd in the league per 48 minutes (eighth in 2017). When adjusting for position, Westbrook was easily the best defensive rebounder in the league, significantly outpacing Lance Stephenson, Dejounte Murray, and Terry Rozier as the only players in the league who were twice as productive as average.

          On the offensive boards, Westbrook was 39th in the league in terms of Net Offensive Rebounding Value. The data does reflect a real difference in Westbrook’s own performance (his value dropped from 1.99 boards per 48 minutes in 2017 to 1.73 per 48 minutes in 2018), though he was still second among offensive number 1’s to only Giannis Antetokounmpo. For 2018, Russell Westbrook ranked 19th in the league in Position-Adjusted Net Offensive Rebounding Value per 48 Minutes, a measure by which he placed 57th in the league in 2017 and 14th in the league in 2016.

          Russell Westbrook’s Total Net Rebounding Value of 446.5 ranked ninth in the league, though he was only 41st in the league per-minute. His overall per-minute performance, then, was lower than in 2017, though consistent with 2016. He placed third in the league in Position-Adjusted Total Net Rebounding Value per 48 Minutes, his highest finish among the seasons under consideration.

Mary Altaffer/Associated Press

          Westbrook was the Thunder’s best defensive rebounder, both overall and per 48 minutes (though detractors will note that he ranked behind Enes Kanter, now of the New York Knicks, in Defensive Rebounding Value per 48). He was the Thunder’s second-most prolific offensive rebounder behind Steven Adams (leaguewide, Westbrook also notably ranked behind former teammates Enes Kanter and Domantis Sabonis). In terms of Total Net Rebounding Value, Westbrook led the Thunder by a margin of 20 rebounds over Steven Adams. Again, detractors will be quick to point out that Enes Kanter bested Westbrook by 37 rebounds’ worth of value. Per 48 minutes, Westbrook was actually OKC’s second-best rebounder, behind Steven Adams. Both Kanter and Sabonis posted superior totals to either of the Thunder’s two leading rebounders in their new habitats.

How About Last Year?

          Last season saw Westbrook achieve the feat of averaging a triple double for the third consecutive season. Relative to his previous performance, Westbrook’s offensive rebounding performance dipped considerably. For the 2018-19 season, he ranked 102nd in the league in total Offensive Rebounding Value, and only eighth among his position group (offensive number 1’s). His per-48-minutes numbers on the offensive glass were pedestrian, placing him in the middle of the league. His position-adjusted value per 48 minutes ranked Russ number 82 among all qualifiers. All of these measures fell far below his previous standard. They also placed well below teammate Steven Adams and ex-teammate Enes Kanter, who were two of the top seven offensive rebounders in the league.

Who got this rebound?
(Photo by Steve Dykes/Getty Images)

          Since we know that Westbrook still averaged double-figure rebounds per game, we would expect to see a corresponding improvement on the defensive boards to offset his decline on the offensive boards. For 2019, Westbrook ranked eighth in the league in total Defensive Rebounding Value, far surpassing all other players of his size and role and finishing within 70 rebounds of all the big men above him except for Andre Drummond. His rate of 6.06 net defensive rebounds per 48 minutes ranked 14th in league, once again far outstripping the normal performance levels associated with a perimeter defender. Among qualifying players, no one bested Westbrook’s mark of 2.96 times more rebounding value than his position average per 48 minutes. In fact, only Kawhi Leonard, DeAndre Jordan, and Jayson Tatum registered above 2.0 on this scale.

          Russell Westbrook finished 19th in the league in Total Net Rebounding Value for 2018-19, ironically falling only a few rebounds behind Enes Kanter (whose playing time was limited with the New York Knicks early in the season) and Steven Adams. Per minute, he was only the 49th best rebounder in the league. When adjusted for position, we find that Westbrook landed at 15th in the league, worth 2.11 times as many rebounds per 48 minutes as the average player at his offensive and defensive positions. We may note that Enes Kanter far exceeded this level, accumulating three times as much rebounding value as the average player in his role would have done.

So Does Russ Steal Rebounds?

          Surveying four years’ worth of evidence, I believe it is fair to say that Russell Westbrook has been an elite rebounder for his position for a large portion of the time period. In spite of taking on a more significant offensive load following the departure of Kevin Durant, Westbrook still performed at an elite level on the offensive glass. Until a drop this past season, Westbrook maintained a consistently excellent level of performance on the offensive boards.

Jesse D. Garrabrant/Getty Images

Westbrook was much more involved on the defensive glass following the departure of KD, though his counting stats do track fairly well with the actual amount of value he added on the defensive boards. Even in 2018-19, when Westbrook’s overall rebounding performance declined, the decline came at the offensive end. His defensive rebounding numbers remained at similar levels to his previous post-Durant seasons. Furthermore, Westbrook improved his standing in position-adjusted defensive rebounding following Durant’s departure, which may indicate that the difference in value was not entirely the result of stat-stuffing.   

          If you want evidence to prove that Russell Westbrook doesn’t pad his stats, you needn’t look any further. Some people, however, will argue that Westbrook’s defensive rebounding production came partially at the expense of his Thunder teammates, and that the slight decrease in offensive rebounding value is evidence of this; Westbrook didn’t actually rebound better in 2017, 2018, and 2019 than he did before the triple double era, he simply rebounded more by taking rebounds that would normally have been collected by the team’s bigs.

          Westbrook was only the indisputable best rebounder on his team in two of the four seasons surveyed: in 2016, he ranked third on the Thunder behind Enes Kanter and Kevin Durant, while in 2018 he ranked second on the team behind Steven Adams in Total Net Rebounding Value per 48. There may be a case to be made that his rebounds per game numbers overstate his value somewhat in 2017-2019.

In the 2018 season, for example, Westbrook ranked 11th in the league in rebounds per 48 minutes, but ranked significantly lower (again depending on the minimum amount of minutes to qualify) in actual value added per 48 minutes. Somewhat similarly, the 2019 season saw Westbrook clock in at 59th in the league in per-minute rebounding, so that his noticeable per-game numbers are partially a representation of his having played a lot of minutes. His counting stats and value track much more closely in 2017 than in either of the subsequent seasons, so it might even be advisable to opt for a nuanced position which admits of some stat-padding by Russell Westbrook but locates most or all of the stat-padding in the 2018 and 2019 seasons (thereby preserving the historical uniqueness of his 2017 MVP campaign).

If you’re looking for evidence that Russell Westbrook does steal rebounds from his teammates, it’s not difficult to find. Since the trade that sent Westbrook to the Houston Rockets last summer, Thunder center Steven Adams’ Defensive Rebound Percentage has skyrocketed above 25% – not unusual for a high-quality center, but far in excess of his percentages during Russell Westbrook’s triple double seasons (15.4%, 13.9%, 14.8%). While we have made mention of the factors which might explain Enes Kanter and Domantis Sabonis getting more rebounds after moving to a team without Russell Westbrook, none of these factors apply with Adams. Steven Adams was well-established as a good rebounder before Russell Westbrook’s triple double seasons, ranking in the top 20 of the league in both total offensive rebounds and Offensive Rebounding Percentage in 2014-15 and 2015-16. He already had enough experience to make it doubtful the changes were the result of skill improvement. The mist likely cause of his explosion in defensive rebounding is that … he doesn’t play on the same team as Russell Westbrook anymore.

So, you can argue that Russell Westbrook steals rebounds from his teammate, or that he doesn’t. You can argue that Westbrook’s rebounding numbers are inflated; on the contrary, you can argue that Westbrook is truly an elite rebounder. Which side you take depends on which evidence you find more persuasive, as well as on how you choose to weight the evidence. Statistics, as the saying goes, are merely opinions stated as facts.


One thought on “Can You Give Me Some Stats to Prove that Russell Westbrook Steals Rebounds?

  1. Excellent article! I like the idea of focusing on contested vs. uncontested rebounds: that’s a new one for me, and it’s terrific! I’m uncomfortable in valuing O rebounds over D rebounds (I’ve seen 3-1 weighting by some analysts) because I believe that the only thing stopping an O rebound is a D rebound. It’s like in football: if you place a high value on pass rushers, then you need to place an equally high value on pass blockers. Contested vs. uncontested moves past that debate, thankfully. I had previously read that Adams had been shown to “give up” the most rebounds in league to teammates (primarily to Russ). If I’ve read it right, this article supports that while at the same time acknowledging Russ’s good rebounding relative to position.


Leave a Reply